A new large-scale study of European adults suggests that, on average, being religiously educated as a child is associated with slightly poorer self-rated health after the age of 50. The research, published in the journal Social Science & Medicine, also indicates that this association is not uniform, varying significantly across different aspects of health and among different segments of the population.
Past research has produced a complex and sometimes contradictory picture regarding the connections between religiousness and health. Some studies indicate that religious involvement can offer health benefits, such as reduced suicide risk and fewer unhealthy behaviors. Other research points to negative associations, linking religious attendance with increased depression in some populations.
Whole thing was an interesting read thanks for sharing. I will say, the article pretty clearly doesn’t support a “religion is poison” thesis. But rather a “religious upbringing is, on average, associated with poorer mental health later in life” with a whole bunch of additional asterisks and details, which is maybe a little more grounded of a way to interpret science generally speaking.
However, the model also identified a smaller portion of individuals for whom the association was positive, suggesting that for some, a religious upbringing was linked to better health outcomes. This variation highlights that an average finding does not tell the whole story.
Future research could build on these findings… …More detailed measures of religious education could also help explain why the experience appears beneficial for some health domains but detrimental for others.
Apparently they used a type of machine learning model to analyze the dataset. I’m not familiar enough to know if there are shortcomings to that approach (as one would expect with gen ai, but machine learning statistical analysis seems less likely to share those faults). Relevant context from the article:
The researchers employed an advanced statistical method known as a causal forest approach. This machine learning technique is particularly well-suited for identifying complex and non-linear patterns in large datasets.
All this with an average of -0.10 ( -2%) difference, for a factor that is almost impossible to isolate even across a large sample. Even if the results were conclusive, I’m not sure what self reported perception of health proves exactly. For example, if Donald Trump filled out the survey and reported his health as all 5s across the board, how do you interpret that?
Whole thing was an interesting read thanks for sharing. I will say, the article pretty clearly doesn’t support a “religion is poison” thesis. But rather a “religious upbringing is, on average, associated with poorer mental health later in life” with a whole bunch of additional asterisks and details, which is maybe a little more grounded of a way to interpret science generally speaking.
The original paper for those interested https://doi.org/10.1016/j.socscimed.2025.118210
Apparently they used a type of machine learning model to analyze the dataset. I’m not familiar enough to know if there are shortcomings to that approach (as one would expect with gen ai, but machine learning statistical analysis seems less likely to share those faults). Relevant context from the article:
All this with an average of -0.10 ( -2%) difference, for a factor that is almost impossible to isolate even across a large sample. Even if the results were conclusive, I’m not sure what self reported perception of health proves exactly. For example, if Donald Trump filled out the survey and reported his health as all 5s across the board, how do you interpret that?